⚛️ Physics

🔬 ICLR2026 · 2 paper notes

Feedback-driven Recurrent Quantum Neural Network Universality

This paper establishes the first quantitative approximation error bounds and universality proofs for feedback-based recurrent quantum neural networks (RQNNs), demonstrating that RQNNs can approximate arbitrary fading memory filters with a linear readout layer while requiring only \(\lceil\log_2(\varepsilon^{-1})\rceil\) qubits — growing logarithmically with precision — and are thus free from the curse of dimensionality.

Sublinear Time Quantum Algorithm for Attention Approximation

This paper proposes the first quantum data structure with sublinear time complexity in sequence length \(n\) for approximating row queries of the Transformer attention matrix. The preprocessing time is \(\widetilde{O}(\epsilon^{-1} n^{0.5} \cdot \text{poly}(d, s_\lambda, \alpha))\) and each row query takes \(\widetilde{O}(s_\lambda^2 + s_\lambda d)\), achieving a quadratic speedup over classical algorithms with respect to \(n\).